MrFewkes - 8:58 am on Jul 19, 2010 (gmt 0)
I am with you as far as the post above goes. It may seem simple - but I looked at LSI with kind of a simple mind - like this....
"If a page wants to rank for the phrase ALBERT EINSTEIN - then it better mention relativity, energy, mass, light etc."
I have a program which crawls sites and provides me with those 4 words (for example that is) - as I dont know them without prior knowledge of einstein for example again.
Then the program plots a graph of page usage of those words for each site in the top twenty. The graph thins out as the words appear less frequently in pages down the serps - the graph showed that the higher pages almost always contained a high percentage of the words which were common to the top 20 docs in the serps.
I think I am correct in assuming therefore that google couldnt implement this on a large scale (as you say due to computational constraints) - having written the code to do it against a few serps I can understand why.
I came to the conclusion that (in theory at least) if I created enough documents wanting to rank for "Computer Screen" - and on each of those documents I placed the word "Treehouse" (unrelated word in natural documents) - then effectively I could make google only ever rank pages for computer screen which contained also the word treehouse. More to it but thats the general gist of it.
So, anyways, back to it, they have done something lesser which is phrase based - that would require less cross-referencing for certain on a massive scale - once a "fixed" DB was acheived.
Hmm - im waffling now.
"So if you hope to rank well for a given phrase, then the presence of a few related phrases on the page might help."
Tedster - I hope so - because in this instance "might help" is one thing - but as a result of this thread - I am actually relying on the implementation of phrases totally stomping on the old style stuff in order to give me a movement.....as as I say - my pages just seem "stuck".